Uncorrected proofs for review only

Transcription

1 0 0 If Drug Treatment Works So Well, Why Are So Many Drug Users in Prison? Harold Pollack, Peter Reuter, and Eric Sevigny Drug use and drug sales play central roles in the history of American crime. One cannot discuss crime in America in the 0s without reference to the heroin epidemic nor in the s and early s without reference to powder and crack cocaine. The highly punitive regime in place now for drug offenders is largely a response to the association of these epidemics with crime waves. Yet these drug epidemics unfolded a long time ago, with apparently low rates of initiation into drug dependence in recent years. From the vantage point of 0, one might think that these drugs no longer matter much, and that dependent drug use plays a smaller role in crime and criminal justice policy than it did ten or twenty years ago. In fact, however, data from many sources (some described later) indicate that those arrested or incarcerated within the American criminal justice system remain heavily involved in the consumption of illicit drugs. There is abundant evidence that inducing criminal offenders to halt or reduce their substance use would reduce crime. Policymakers, researchers, and advocates have long argued that broader provision of substance abuse treatment could reduce the number of Americans behind bars. In fact, however, a major component of the relentless growth in the US incarcerated population over the last thirty- five years has been the rising Harold Pollack is the Helen Ross Professor at the School of Social Service Administration and faculty chair of the Center for Health Administration Studies at the University of Chicago. Peter Reuter is a professor in the School of Public Policy and in the Department of Criminology at the University of Maryland. Eric Sevigny is an assistant professor in the Department of Criminology and Criminal Justice at the University of South Carolina. We thank Holly Nguyen for capable research assistance. We thank Jonathan Caulkins, Steven Raphael, and the conference organizers for excellent comments.

2 Harold Pollack, Peter Reuter, and Eric Sevigny 0 0 number of people imprisoned for drug offenses, (Blumstein and Beck ) a figure that rose from,000 in to,000 in 00 (Caulkins and Chandler 00). Most of this incarceration burden falls on people who were involved in supplying drugs, albeit sometimes in minor roles (Sevigny and Caulkins 00). In addition, a large number of those incarcerated for both drug and nondrug offenses appear to satisfy screening criteria for drug use disorders. Many are dependent on cocaine, heroin, or methamphetamine; there is reasonable evidence that their drug use has a causal role in their criminality (MacCoun, Kilmer, and Ritter 00). Substance abuse treatment provides a highly imperfect response to these problems. During any given treatment episode, the typical client is likely to continue some level of substance use. Relapse is the norm rather than the exception as a treatment outcome. Even so, at the individual level, there is compelling evidence that treatment markedly reduces both drug use and related criminal offending. Imperfect treatment works. This is just as well, since it describes the treatment we have. An array of programs have developed over the last twenty years based on this evidence, and more broadly on the well- documented premise that reducing drug use leads to large reductions in the individual offender s crime rate. The list of programs includes drug courts, other forms of diversion from the criminal justice system into treatment (e.g., Proposition in California), intensive supervision probation, and in- prison treatment. All these aim to reduce the extent of criminality among those who have already developed drug abuse or dependency by encouraging/ coercing offenders into treatment. A substantial research literature shows that treatment does reduce both drug use and associated criminal activity. In addition to such programs, there is growing recent interest in coerced abstinence or mandated desistance interventions, whereby drug- involved offenders under criminal justice supervision in noncustodial settings (parole, probation, and pretrial supervision) are subject to short, immediate, and graduated penalties for detected drug use (Kleiman 00). Despite this array of efforts, there has been no decline in the incarceration of drug users for either drug offenses or for other criminal activities. The number incarcerated for drug offenses has increased every year since (Caulkins and Chandler 00). We show later in this chapter that the number of state prisoners with drug problems also increased substantially from to 00, extending analyses of The National Center on Addiction and Drug Abuse at Columbia University (CASA) (Belenko et al. 00) and Mumola and Karberg (00). We find strikingly similar patterns within the increasingly important population incarcerated in local jails. Both of these findings are rather surprising, since the number of individuals with expensive illegal drug habits who are not incarcerated was estimated to have declined in the period to 000, the most recent years for which a published estimate is available, (Office of National Drug Control Policy

3 If Drug Treatment Works So Well, Why Are So Many Drug Users in Prison? ) and there are some indicators that the decline may have continued. This would suggest that there are fewer sellers as well as fewer users to lock up. Why has the United States achieved such limited success in getting criminal offenders to curtail their drug use? Put slightly differently, why is it so difficult to replicate at the population level the substantial reductions in drug use and criminal offending that treatment appears capable of achieving for individual offenders? Why aren t more offenders in treatment? And why have diversion programs such as California s Proposition and drug courts proved relatively disappointing in achieving their stated goals? We hypothesize that there are two main reasons for the continued large numbers of drug users flowing into, and remaining within, the correctional system: First, eligibility criteria for diversion programs, particularly for drug courts, are restrictive. Although the various programs are effective and even cost- effective in serving the specific clients they recruit, they make only a small contribution at the population level. The diverted offenders are at low risk of going to prison or even jail (following sentencing, as opposed to pretrial) in the absence of the drug court intervention. Given limited capacity and the relatively low- risk populations actually served, the currently deployed model of drug courts is unlikely to notably reduce prison populations. A second, related pattern also hinders the effectiveness of these interventions. There is a systematic mismatch between sentencing practices and actual criminal careers among drug- involved offenders. As individual criminally active drug users get older, the system increasingly treats them harshly for each successive offense. They have longer criminal histories, longer records of unsuccessful treatment, and worse employment histories. Thus, not only are they less eligible for diversion programs, these offenders also receive longer sentences, increasing the share of the incarcerated population with drug problems. The empirical contribution of this chapter primarily concerns the first of these conjectures. In particular, we examine what share of those currently incarcerated would have been eligible for drug courts with the least restrictive entry criteria. We have not been able to find data that allows testing of the effect of the potentially lengthening criminal careers of dependent drug users. To test the hypothesis about the ineffectiveness of drug courts to reduce the size of the incarcerated drug- offending population, we make use of the Survey of Inmates in State Correctional Facilities (SISCF) and the Survey of Inmates in Local Jails (SILJ), two Bureau of Justice Statistics (BJS) occasional surveys. Both provide self- reports on, inter alia, criminal activity and substance use from nationally representative samples of inmates; the Prison survey has been conducted six times between and 00 (with federal

4 Harold Pollack, Peter Reuter, and Eric Sevigny 0 0 inmates surveyed only in the and 00 studies), while the Jail survey has been conducted six times between and 00. We find that, indeed, very few of those entering state prison in 00 or jail in 00 would have been eligible for diversion through state courts. That this is true for local jails is much more surprising than the prison finding. This pattern provides a reminder that, even late in the incarceration boom, it is not so easy to get incarcerated, conditional on arrest. There are two reasons for the findings about drug courts. First, many entering prison and jail (whether drug users or not) were on supervised release (parole or probation) at the time of their latest arrest, which automatically made them ineligible for most drug court interventions. Second, and more interestingly, most of those who were arrested de novo and who had drug use patterns making them potential clients for drug court, had long, relatively serious criminal records that would have made them ineligible under current conditions. Drug use itself may lead to more intense or longer criminal careers. Moreover, many of those dependent on expensive drugs (cocaine, crack, heroin, and methamphetamine) became drug users a long time ago. These populations are aging, which is not true of nondrugusing criminal offenders. In effect, what we are seeing is two distinct trends in the incarcerated population, separated by drug use. We also present three other policy- relevant descriptive findings: First, it is useful to compare the number of dependent drug users entering treatment with the number entering prison. Both in and 00, these figures are approximately comparable; the United States is locking up about as many drug addicts as it is treating, a troubling observation about the nation s drug policies. Second, there are indications that drug dependence is less prevalent among younger offenders than in cohorts that are twenty years older. Absent a new drug epidemic or a newly invigorated drug war, there is a predictable end in sight to the growth of drug- related prisoners. Third, for drug using prisoners, the probability of a violent offense declines sharply with age after thirty- five. This last observation leads us to our principal policy suggestion, which needs further investigation. Diversion programs of all kinds require substantial redesign if they are to contribute to a reduction in the incarcerated population. Experienced drug users, who account for an increasing share of drug- related crime, are not attractive (or eligible) candidates for many current efforts. However, if one is willing to take a very long- term social welfare perspective, it may be worth introducing courts specifically designed for the long- term user. Our finding that aging drug users commit relatively few violent crimes is helpful here. The risks associated with treatment- oriented community supervision of older offenders are therefore less than one likely encounters in younger drug- using cohorts. The chapter begins with three review sections. Section. describes the

5 If Drug Treatment Works So Well, Why Are So Many Drug Users in Prison? 0 0 changing patterns of drug misuse in the United States over the last forty years, which is necessary to understand the challenge now facing the criminal justice system. Section. follows with a review of what is known about the effectiveness of drug treatment in reducing crime at the individual level. Section. briefly discusses interventions aimed at diverting drug- involved offenders from incarceration, such as drug courts, Proposition (the largest diversion program in operation, even though it is restricted to California alone), and coerced abstinence/ mandated desistence in Hawaii. Section. presents our empirical analysis of the surveys of jail and state prison inmates, showing the limited potential impact of drug courts under current eligibility rules. Section. presents our conclusions.. Background: The Changing Demography of Drug Misuse The dynamics of drug- related incarceration in the United States should be examined in light of broader societal trends in drug use and dependence over the last forty years. The characteristics of the drug using population, particularly those dependent on expensive drugs, has changed in ways that complicate the task of keeping criminally active drug users out of prison... Drug Epidemics The nation has experienced four major drug- specific epidemics in that period; heroin (ca. to ), cocaine powder (ca. to ), crack cocaine (ca. to ), and methamphetamine (ca. to 000). In an epidemic process, rates of initiation rise sharply as new and socially contagious users of a drug initiate friends and peers, a model first well developed by Hunt and Chambers (). In the case of heroin, there is much evidence of a sudden elevation of initiation rates during the late 0s and early 0s, followed by a rapid incidence decline over the 0s and s (Kozel and Adams ; Rocheleau and Boyum ). A study of an early s sample of street heroin users also found evidence of sharply peaked initiation rates in the late 0s and early 0s. For cocaine powder the rise was similarly rapid but decline was not so pronounced as with heroin (Everingham and Rydell ). For crack cocaine the epidemic was still later, starting between about and, depending on the city (Cork ). Caulkins et al. (00) reported estimates of annual cocaine initiation using the National Household Survey on Drug Abuse (NHSDA) and a variety of methods; all show a peak in followed by a decline of two- thirds in the next five years. A new class of epidemiologic models has been developed by Caulkins and collaborators (Caulkins et al. 00; Caulkins 00), which use diverse data to document the long trajectory of drug epidemics. After the peak, the initiation rate does not return to its original zero level but falls to a rate well below the peak. Under reasonable assumptions, the result is a flow of

6 Harold Pollack, Peter Reuter, and Eric Sevigny 0 0 new users who do not fully replace those lost through desistance, death, or incarceration. Thus, the number of active users declines gradually over time. Moreover, the drug- using population ages with corresponding changes in the health, employment, and crime consequences of substance use. Some evidence for this characterization can be seen in the changing characteristics of drug users in TEDS (the Treatment Episode Data System) that includes data on admissions to treatment programs that receive federal funds. We do not report changes in the National Household Survey on Drug Abuse/ National Survey on Drug Use and Health (NHSDA/ NSDUH) because these include so few dependent users. For TEDS we are able to compare the admission cohort of with that of 00; these two years are the earliest and latest for which detailed data are available. By all but the methamphetamine epidemics had run their courses but the cocaine and crack epidemics were relatively recent, so many of the users showing up for treatment were still young adults. Figure., computed using and 00 TEDS data, displays changes in the age distribution of adult clients admitted into substance abuse treatment who reported cocaine- related disorders. In the data, percent of clients were under the age of thirty. By 00, that figure had dropped to percent. The fraction of clients over the age of forty rose from percent to percent over the same period. This was not the consequence of an epidemic of new use among older individuals; rather, it represented the aging of those who were caught in the earlier epidemics. We observed a more complex pattern within the population of admitted heroin users. As shown in figure., the over- forty- five population displayed a similar pattern to that found in the population of cocaine users. Yet there was also a substantial population of admitted heroin users below the age of thirty. The Drug Abuse Warning Network (DAWN) shows similar patterns of the aging of cocaine and heroin users appearing in emergency departments or as overdoses examined by medical examiners through 00. The result of this epidemiology is that the demography of drug misuse changed substantially between the early s and the 000s. The average age of drug users increased markedly, with a more diverse set of primary drugs of abuse. These data suggest that current service utilization reflects the long- term reverberation of specific epidemics of cocaine and heroin use in the United States. They also matter for the criminal justice system.. For example, in 000, self- reported prevalences among NHSDA respondents imply that. million individuals had used cocaine in the previous month. By comparison, more broadbased estimates that included ADAM estimated a total of approximately two million that met the more stringent requirement of having used the drug more than eight times in the thirty days prior to the interview. The differential for heroin was similar (Boyum and Reuter 00, ).

7 If Drug Treatment Works So Well, Why Are So Many Drug Users in Prison? 0 0 Fig.. Age distribution of TEDS cocaine admissions, and 00 Fig.. Age distribution of TEDS heroin admissions, and 00 The only published estimates, distributed by the Office of National Drug Control Policy in 00, of the numbers of dependent cocaine and heroin users cover the period to 000 (Office of National Drug Control Policy 00). Figure. presents these figures, which rely heavily on Arrestee Drug Abuse Monitoring (ADAM), showing a substantial decline, about one- third for each drug, over these twelve years. Both the data and the estimation methodology are weak, as indicated by the frequent adjustment in single year estimates in successive series published by the same research group over the period to 000. For example, ONDCP s immediate preceding version of the estimates had shown an increase in heroin use in the early s, followed by a rapid decline (Office of National Drug Control Policy 000). Some of the decline in these estimates may represent the consequence of increased incarceration, since those in prison are not eligible for the ADAM

8 Harold Pollack, Peter Reuter, and Eric Sevigny 0 0 Fig.. Number of chronic cocaine and heroin users (in thousands), 000 Source: What America s Users Spend on Illicit Drugs 000 (ONDCP 00). sampling frame. Assume for the purposes of a rough calculation that the share of state prison inmates who would be classified as cocaine- or heroindependent prior to entering prison rose from percent of the,000 in to 0 percent of the,,000 in 000. That would have removed roughly 00,000 dependent cocaine and heroin users from the pool on which these estimates are based. Other trends may also account for some of the observed decline. For example, almost 00,000 injection drug users have died of HIV/ AIDS. Although these trends are important for many reasons, they account for less than one- third of the total decline (from. million in to. million in 000). The best interpretation of the available data is that the number of individuals dependent on or abusing expensive drugs has been declining for a long period for a variety of reasons. The population of such users has aged, presumably reducing their involvement in violent crime.. Drug Treatment Though the research has been critiqued by the National Research Council, (Manski, Pepper, and Petrie 00), a substantial body of evidence indi-. The calculation is done only for state prisoners because (a) jail inmates serve short terms on average and are eligible within the year for rearrest and ADAM inclusion, and (b) federal inmates include a large fraction of nonresident offenders.

9 If Drug Treatment Works So Well, Why Are So Many Drug Users in Prison? 0 0 cates that substance abuse treatment is associated with large reductions in drug use and crime, especially during the period in which the individual drug user is in treatment. United States and British observational cohort studies document the strong association between treatment receipt and increased employment, improved health outcomes, and reduced criminal offending. For example, Godfrey, Stewart, and Gossop (00) reported two- year outcome data for drug users enrolled in the British National Treatment Outcome Research Study (NTORS). Expenditures on substance abuse treatment for these individuals totaled. million British pounds. Economic valuation of treatment- associated crime reduction totaled. million pounds for the same group. A four- to five- year follow- up indicated reductions in the frequency of heroin, street methadone, and benzodiazepines (Gossop et al. 00). Crack cocaine and alcohol use were not significantly different after four- to five- years from the corresponding values at intake. Analyzing the same data, Gossop et al. (00) found substantial reductions in acquisitive, drug selling, and violent crimes. Crime reductions were associated with reduced regular heroin use, simple aging, and living in stable housing. Similar results were observed with US data collected by the Drug Abuse Treatment Outcome Study (DATOS). Hubbard et al. () reported that clients able to remain in long- term residential treatment for at least six months exhibited a 0 percent reduction in illegal activity and a percent increase in full- time employment. Koenig et al. (00) and Ettner et al. (00) observed similar patterns among treatment clients in Cuyahoga County, Ohio, and California, respectively. Both of the latter papers reported strongly positive net benefits from treatment, with reductions in criminal offending accounting for the majority of the observed economic benefit associated with treatment intervention. Prendergast and colleagues provide one widely cited meta- analysis of these effects (Prendergast et al. 00). These authors examined results from seventy- eight studies completed between and. Twenty- five of these analyses also examined crime outcomes; forty- six featured randomized study designs. These authors found that treatment was associated with reduced drug use (effect size 0.0) and reduced crime (effect size 0.). In examining the impact of treatment on crime, reduced substance use appeared to be the critical mediating variable, with reduced substance use inducing lower rates of acquisitive crime. The average age of participating drug users was the only significant predictor of effect size, with treatment having a larger absolute impact in reducing crime among young adults (who are the most criminally active) than among older drug users. Some of the strongest findings for outpatient treatment arise in the arena of methadone maintenance therapy. For example, Amato et al. (00) found that methadone maintenance therapy reduces criminality by as much as

10 Harold Pollack, Peter Reuter, and Eric Sevigny percent. The findings from a long- term cohort study of heroin users by Hser and colleagues finds similar results for a particularly recalcitrant heroin- using population (Hser et al. 00). Retention and treatment outcomes among opiate users appear sensitive to specific quality measures (D Aunno and Pollack 00). For example, methadone maintenance clients were markedly more likely to remain in treatment (AOR.) when methadone doses exceeded sixty mg/ day (Bao et al. 00). Many studies of opiate substitution therapy (OST) indicate a strong negative correlation between treatment engagement and retention on the one hand, and criminal offending on the other. Campbell, Deck, and Krupski (00) examined arrest rates among Washington state opiate users. These authors found significantly reduced probability of arrest among treatment participants. Burdon, Messina, and Prendergast (00) found quite similar results among California offenders participating in prison aftercare. Given these strong relationships, both state and federal prisons sought to increase treatment provision to drug- involved offenders (Grella et al. 00; Taxman, Perdoni, and Harrison 00). Most such services are low- intensity education and counseling services, which probably have a limited impact on criminal offending or drug use. More intensive residential modalities have also been implemented in prison, with greater evidence that treatment participants achieved better outcomes than comparison group members (Prendergast et al. 00). However, because motivated individuals are more likely to enter and remain in treatment, many of the observed differences between treatment and comparison groups probably reflect favorable selection into treatment. Given the possibility indeed the reality of strong selection effects, randomized trials are especially important in evaluating the causal impact of treatment interventions. In one recent Australian study, Dolan et al. (00) compared reincarceration, treatment mortality, and hepatitis C infection rates among opiate- dependent prison inmates randomly assigned to methadone maintenance and to a control group. Members of the treatment group displayed lower incidence of hepatitis C. Yet assignment to the treatment group appeared to provide little benefit in terms of long- term treatment retention. In several recent papers, Gordon et al. (00) and Kinlock and colleagues examined drug and crime outcomes of heroin- dependent Baltimore prisoners who were randomly assigned to methadone maintenance or a control- group counseling intervention (Kinlock et al. 00; Kinlock et al. 00; Kinlock et al. 00). Offenders offered methadone maintenance shortly after release were significantly less likely to use heroin/ cocaine or engage in criminal activity compared to those assigned to the control group. A randomized trial by McMillan et al. (00) yielded less favorable findings. Offering opiate- dependent inmates methadone maintenance within the

11 If Drug Treatment Works So Well, Why Are So Many Drug Users in Prison? 0 0 jail setting appeared to confer little benefit absent an effective postrelease intervention. Outside the arena of opiate substitution therapy, the strongest evaluation results arise in establishing the benefits of therapeutic communities. For example, McCollister et al. (00) conducted a five- year follow- up study examining the Amity in- prison therapeutic community and an accompanying Vista aftercare program for criminal offenders in southern California. The average cost of addiction treatment over the baseline and five- year follow- up period was $,0 for the intervention group and $, for the control group. However, the treatment group experienced eighty- one fewer incarceration days than was observed within the control group. This percent reduction in incarceration more than offset the additional costs of the relatively intensive intervention. Evaluations of outpatient drug- free interventions yield more mixed results. In the case of cocaine, a meta- analysis of research on interventions aimed at dependent users of a variety of drugs, few of whom were in methadone maintenance, found that those in treatment were about 0 percent more likely to have positive outcomes with respect to criminality than those who did not enter treatment (Prendergast and Burdon 00). Even though most who enter treatment will relapse to drug use and/ or fail to complete their treatment, it is still true that treatment can make a large difference in the lifetime drug use and criminality of a dependent user. These large differences in criminal offending lead to correspondingly large impacts in cost- benefit analyses of substance abuse treatment. Substance abuse treatment is associated with many economic benefits. Yet crime reduction is consistently the largest single component of the economic benefit of treatment (Dismuke et al. 00; Sindelar et al. 00; French et al. 00; McCollister and French 00). Indeed, the economic benefits of treatmentassociated crime reductions are often larger than all other estimated benefits combined. The economic valuation of treatment- related crime reduction frequently exceeds, by itself, the entire cost of providing substance abuse treatment. In one prominent analysis of cocaine- dependent clients, Flynn et al. () examined treatment clients self- reported crime before and after treatment, finding that the economic value of an associated reduction in crime far exceeded the associated treatment costs. Such findings are doubly striking because most studies in the empirical treatment literature understate the true social benefits associated with reduced crime. Most studies consider the tangible costs of crime its direct costs to victims and to the health care and the criminal justice systems. The tangible cost approach provides a valuable lower bound to the benefits of. Analysis that accounted for the social benefits of averted crime would likely find even more striking benefits of substance abuse treatment.

12 Harold Pollack, Peter Reuter, and Eric Sevigny 0 0 Table. Treatment admissions for cocaine, heroin, marijuana and methamphetamine, TEDS and 00 (N) (%) 00 (N) 00 (%) Cocaine,0 0, Heroin,, Marijuana,, Methamphetamine,, Note: Each admission is classified according to primary drug of abuse but may involve polydrug abuse. Reported percentages are based on treatment admissions for all substances, including alcohol. crime reduction. However, such costs are a small fraction of the overall social costs of crime (Rajkumar and French ). Flynn et al. () cite tangible costs of $,0 per burglary. By contrast, Cohen et al. (00) obtain a per burglary cost estimate of $,000 using contingent valuation methodologies that capture a broader range of crime consequences and societal preferences (Cohen et al. 00). Basu, Paltiel, and Pollack (00) perform a (nonexperimental) prepost analysis of US treatment data from the National Treatment Improvement and Evaluation Study (NTIES) that illustrates the importance of these valuation measures. Using conservative econometric specifications that were biased against a finding of treatment effectiveness, these authors show that the monetized value of treatment- related reductions in armed robbery more than offset the cost of the entire treatment intervention. This finding is especially striking when one considers that less than percent of NTIES respondents reported committing an armed robbery in the year before treatment admission. Moreover, only percent of these robbery offenders reported ever being arrested for using a weapon or force to steal from a victim. In terms of absolute numbers, substance abuse treatment providers serve a large and diverse population of substance users. As shown in table., the number of individuals in drug treatment for cocaine or heroin abuse has risen slightly since ; for example, TEDS data indicate admissions of,000 in for heroin, compared to,000 in 00. Given that the estimated size of the population of dependent users has, if anything, shrunk, this indicates that the treatment fraction has increased. The figure for methamphetamine admissions almost tripled during the same period.. Drug Use and Crime The criminally active population continues to show high rates of drug misuse, another indication that treatment has, at the population level, failed to reduce the connection between crime and drug use. For this population, ADAM provides the major source for insights into the connection between crime and drug use. The ADAM survey includes

13 If Drug Treatment Works So Well, Why Are So Many Drug Users in Prison? 0 0 data on drug use, both through interview and urinalysis, from a sample of arrestees in a number of counties around the country. Prior to we must rely on the Drug Use Forecasting system (DUF), a statistically more primitive version of ADAM but one that turns out to provide data of comparable quality. When ADAM was operating most broadly, from to 00, the data were collected in thirty- five counties. Data were not collected from 00 to 00, and since 00 have been collected in only ten counties (Office of National Drug Control Policy 00). Thus, ADAM provides an incomplete depiction of drug use among the arrested population nationally, particularly since 00; city- level comparisons are more appropriate for comparing trends over time. The most recent ADAM results (for 00) show that use of cocaine, heroin, and methamphetamine continues to be common among arrestees in most cities. The percentage testing positive for cocaine varied between a high of percent and a low of percent. Figures for heroin were lower, but were still as high as percent for Chicago, twice as much as the next highest city. For our purposes the more relevant comparisons are between and 00, the era covered by the two inmate surveys we analyzed. The DUF system started collecting data in in just twenty- one cities (Wish and Gropper ). More complete and consistent data are available from Washington, DC, which has collected urinalysis on all adult arrestees since and on all juvenile arrestees since. The adult data (figure.) show that the percentage of all arrestees testing positive for any drug excluding Fig.. Arrestees testing positive for various drugs in the District of Columbia, 00 Source: Pretrial Services Agency.

14 Harold Pollack, Peter Reuter, and Eric Sevigny 0 0 Fig.. Drug use among arrestees under age, the District of Columbia, 00 Source: Pretrial Services Agency. marijuana declined from percent in to percent in, with a stable trend over the following fifteen years. For juvenile arrestees (figure.) what is striking is how few have tested positive for any drug other than marijuana and PCP since the early s; whereas in percent tested positive for cocaine, that figure has hovered around percent since... Getting Offenders into Treatment As already noted, the insight that drug treatment could reduce both crime and the demands on the criminal justice system has animated policy for a long time, not just in the United States but in other countries. For example, the United Kingdom saw a near doubling of the population in treatment between and 00, largely as a consequence of a large variety of criminal justice diversion programs (Reuter and Stevens 00). We identify here just the major interventions. Drug courts aim to use the coercive power of the criminal justice system, particularly the authority of a judge, to persuade drug- involved offenders to cease their drug use rather than face penalties for violating the terms of their release. Drug court clients are encouraged to seek treatment, and. Most of the late s decline was the result of an abrupt reduction in the percentage testing positive for phencyclidine (PCP), a drug that has been much more prevalent in Washington than any other city, even after the decline.

15 If Drug Treatment Works So Well, Why Are So Many Drug Users in Prison? 0 0 continued participation in treatment may be a condition for staying out of jail. The evaluation literature, though not technically strong, has generally found positive effects on recidivism, the usual outcome measure (Belenko and Peugh 00; Wilson, Mitchell, and MacKenzie 00). Even though the drug court movement is almost twenty years old and over,00 separate programs have been created, (BJA Drug Court Clearinghouse Project 00), a 00 study estimated that only,000 drug- involved defendants were processed in such courts in the middle of this decade; the same study estimated that over one million such defendants entered the criminal justice system each year (Bhati, Roman, and Chalfin 00). Despite the rapid expansion of drug courts, the number of defendants who pass through such programs remain relatively low. This small number of enrollees arises from several factors. Many jurisdictions lack administrative capacity to implement drug courts at- scale. Fifty- two percent of adult drug courts responding to one survey reported they cannot accept some eligible clients due to capacity constraints (Bhati, Roman, and Chalfin 00). Bhati and colleagues estimate the current number of drug court slots at approximately,000. Given this constraint, there are strong administrative and political incentives for drug courts to cream- skim by serving relatively low- risk populations most likely to achieve successful outcomes rather than high- risk populations that would experience the greatest net reduction in criminal offending from drug court interventions. Even if such administrative capacity were available, enrollment would remain sharply constrained by current eligibility restrictions. Despite the pervasiveness of the drug treatment court model, drug courts routinely exclude most of the drug- using offenders. A 00 survey of adult drug courts found that only % of drug courts accept clients with any prior violent convictions. Individuals facing a drug charge, even if the seller is drug- dependent, are excluded in 0% of courts for felony sales and % of courts for misdemeanor sales. Other charges that routinely lead to exclusion include property crimes commonly associated with drug use (theft, fraud, prostitution), and current domestic violence cases (only 0% accept domestic violence cases) (Rossman, Zweig, and Roman 00). An earlier study conducted by the Government Accountability Office () found that only percent of drug courts accept offenders whose current conviction included a violent offense. A study of drug courts in six Washington state counties found substantial variation in eligibility requirements (Cox et al. 00). In King County, for example, only defendants facing drug possession charges were eligible; whereas in Pierce County a long list of property crimes charges were also eligible. Similarly, Florida s Dade County accepts offenders with mainly possession or purchase of a controlled substance charges. Marion County stipulates that eligible offenders must be charged with nonviolent drug offenses, with

16 Harold Pollack, Peter Reuter, and Eric Sevigny 0 0 some drug sale and domestic violence cases considered (Florida Supreme Court Task Force on Treatment- Based Drug Courts 00). Among the seven drug courts in New York City, three accept offenders facing drug sales charges, four do not. Only one court of the seven accepts defendants with nondrug felony charges. Drug courts originally targeted first- time offenders who were arrested for possession or selling to support their habit. Some programs, however, are expanding to include repeat offenders and a few are accepting violent offenders (Porter 00). Nevertheless, programs with flexible eligibility criteria are rare. Table. presents eligibility criteria for four drug courts in major jurisdictions. More difficult to determine are the eligibility rules with respect to substance abuse. Bhati, Roman, and Chalfin (00) report that eligibility based on drug use severity is applied inconsistently % of drug courts exclude those with a drug problem that is deemed too serious, while % reject arrestees whose problems are not severe enough. Almost % exclude those with co- occurring disorders. Even among eligible participants, more than half of drug courts (%) report they cannot accept some clients who are eligible for participation due to capacity constraints (). These eligibility rules seem likely to exclude most experienced users of cocaine, heroin, and methamphetamine. The few cohort studies of cocaine and heroin users (e.g., by Hser and colleagues) show that long- term users have accumulated long histories of convictions for property and violent crimes and that many perhaps most have co- occurring disorders or are polydrug users (Hser et al. 00; Hser 00; Hser et al. 00). Table. Eligibility requirements for four major drug courts County Program type Capacity Eligibility Dade county, FL Adult pretrial,0 No history of violent crime No arrest for drug sale or trafficking No more than two previous felony convictions Brooklyn, NY Postplea NA No prior felony conviction No charges involving drug sale near school No prior felony convictions San Francisco, CA Preplea 0 No drug court failures in previous five years No convictions for sales in previous eight years No current conviction for violent or serious felony Broward, FL Pretrial and, No prior felony or conviction postconviction No current sales/trafficking conviction

17 If Drug Treatment Works So Well, Why Are So Many Drug Users in Prison? 0 0 Estimating the potential effect of relaxing eligibility requirements is a major research challenge. Existing effectiveness findings reflect these tight eligibility requirements. Drug courts choose certain clients, and exclude more serious offenders, in the belief that defendants with longer and more serious criminal histories are more likely to have poor outcomes in drug courts. They may be correct; without evaluations of the effects with these other client groups, the research strategies for making projections are inherently speculative. We take this up in more detail in the chapter s conclusion. California s Proposition provides the largest instance of diversion from the criminal justice system. Under Prop (formally the Substance Abuse and Crime Prevention Act (SACPA)), first- or second- time drug possession arrestees with no record of violent offenses are subject to a drug abuse assessment to determine appropriate referral to a drug treatment program. Parolees or probationers who violate the drug conditions of their release or are arrested for drug possession are also eligible for Prop sentencing to treatment or counseling. Participation in Prop is contingent on pleading guilty to the possession charge. The majority of those arrested for simple possession of marijuana had more attractive legal options, which did not involve pleading guilty and thus did not enter Prop. Passed in 000 by popular referendum, in its fifth year of operation (July, 00 to June 0, 00) it processed,000 individuals, (Urada et al. 00) almost as many as the national drug court movement. Given that Prop is focused on individuals early in their criminal careers, it would appear to have little prospect of reducing prison populations. Yet it appears to have had a sizable effect. For example, Ehlers and Ziedenberg (00) argue that Prop accounted for a decline in the projected California prison population. Whereas the population had been projected to increase from,000 to,000 between 000 and 00, the actual figure ended at only,000. Moreover, the rate of prison commitments for drug possession offenses in California fell from 0 per 0,000 on June, 00 (date of implementation) to per 0,000 four years later. The state has funded a series of detailed evaluations of the effects of Prop (Urada et al. 00; University of California, Los Angeles Integrated Substance Abuse Programs 00). A forty- two- month follow- up of the first wave of Prop arrestees found that the measure substantially reduced the levels of jail and prison incarceration of eligible Prop offenders. The comparison was made between those deemed eligible for Prop in the year before enactment and those who were eligible under Prop in its first year. For state prisons, it appears that offenders who would have been eligible for sentencing under Prop prior to its passage spent 0 days in state prison, whereas those who were sentenced under Prop terms spent only about sixty days in prison. For county jails, the figures were similar; the average length of. This is based on dividing the estimated state prison costs provided in the report by the reported daily cost of a prison stay.

18 Harold Pollack, Peter Reuter, and Eric Sevigny 0 0 time in jail over forty- two months fell from approximately ninety- five days to sixty- five days. The evaluations do not provide information on how much of that difference was accounted for by the initial incarceration spell, and how much was due to subsequences differences in reincarceration. Almost three quarters of those who were processed under SACPA entered treatment. Substantial fractions drop out at various points in the process. The end result is that only one- third (one quarter of the initial intake) were discharged as having completed treatment. That figure is consistent with other studies of outcomes of treatment episodes resulting from criminal justice referrals. Given that Prop clients are under much less threat of reincarceration than those entering through drug court referrals, for example, this is a surprising and encouraging finding, though it underscores the challenge of retaining criminally active populations in treatment interventions. The Prop population has some unexpected characteristics. For example, though this option is only available for first or second convictions on drug possession charges, it is a relatively old population with an average age of. years. Half have never entered treatment before. A substantial percentage have lengthy criminal records, even though this cannot include conviction for a violent offense. Perhaps the inclusion of parolees among Prop eligibles is an important source of the reduction in the prison and jail populations. However, only percent of the clients in the most recent year of the evaluation entered from parole, and that figure had been as low as percent in the first year of the program. Over half of those sentenced under Prop were charged with possession of methamphetamine, a drug associated with high levels of criminality. However, ADAM data show a low prevalence of methamphetamine in most US cities (Office of National Drug Control Policy 00), so the California reductions in incarceration may not generalize to other states. On their face, the Proposition findings are more encouraging than drug courts as a method for reducing drug- related incarceration at the population level. The California assessments consistently suggest that the use of noncriminal penalties has not produced increases in crime rates, either as a result of higher recidivism or of reduced deterrence. Though high- risk arrestees, primarily those with many prior arrests, fare less well than others, treatment- oriented diversion aimed at nonmarijuana possession arrestees may generate a meaningful reduction in total incarceration. There are, however, some concerns that the state- sponsored evaluation, which is complex and not always clear, does not capture all the problems of implementation. For example, Hawken presents an analysis that finds that at the thirty-. The report does not offer a figure for the cost of a day in jail; these calculations assume that it is $.0, a figure cited by Ehlers and Zeidenberg (00). Ehlers and Zeidenberg estimate the number of jail days saved per client per year to be approximately twelve, which is roughly consistent with the thirty days estimated in the above calculation.

19 If Drug Treatment Works So Well, Why Are So Many Drug Users in Prison? 0 0 month mark, arrests for all groups of Prop arrestees (treatment completers, treatment dropouts, and treatment refusers) are higher than for the control group (Hawken 00). Coerced abstinence/ mandatory desistence, a twenty- year crusade by UCLA s Mark Kleiman (Kleiman 00; Kleiman ; Kleiman and Hawken 00), is a program that takes advantage of simple findings from behavioral economics, psychology, and public policy. A large number of offenders are under community supervision at any one time, whether it be pretrial release, probation, or parole. Because they have been arrested or convicted, the government can subject these individuals to random drug tests and indeed does from time to time. Coerced abstinence involves making sanctions certain, immediate, and relatively mild rather than (as is normally the case) random, delayed, and severe. Such interventions have not received widespread evaluation. The small number of existing studies have found that such programs have the predicted effects on recidivism (Harrell, Cavanagh, and Roman ). So far, there have been no efforts to implement them on a large scale. Recently, Hawaii s Opportunity Probation with Enforcement (HOPE) program has implemented the approach for the entire probation population. The results of a random assignment evaluation (Kleiman and Hawken 00; Hawken and Kleiman 00) have been very promising. Very few of those enrolled in the program fail more than twice and the recidivism rates have been dramatically lower than for the probation population previously. For example, only percent of HOPE subjects were rearrested in the twelvemonth evaluation window, compared to percent among those on routine probation conditions (Hawken and Kleiman 00). These results and a clear articulation of the theory underlying the model by Mark Kleiman and others have given this intervention a great deal of political and professional prominence. HOPE- like experiments are being launched in a number of states. It offers the prospect of a large- scale intervention that could be implemented relatively rapidly and without requiring the development of a new expertise in the probation community. However, for those interested in promoting drug treatment as a major intervention to reduce the incarcerated population, it is striking that coerced abstinence does not necessarily involve treatment. Probation officers want their clients to desist from drug use, and this program gives them the tools to motivate and monitor abstinence. Many drug- involved offenders do not satisfy screening criteria for actual dependence. It is unclear that many of the successful clients entered drug treatment programs or that these individuals needed such services. The adverse consequences of a failed urine test have been enough to generate abstinence. Whether abstinence will continue postsupervision is an open question but in making a judgment about the utility of coerced abstinence, it is important to note that relapse is the common experience posttreatment.

20 Harold Pollack, Peter Reuter, and Eric Sevigny 0 0 Table. Results for the HOPE evaluation involving experienced offenders at risk of jail or prison HOPE (%) Control (%) No- shows for probation appointments (average of appointments per probationer) Positive urine tests (average of tests per probationer) New arrest rate (probationers rearrested) Revocation rate (probationers revoked) Incarceration (days sentenced) days days The HOPE evaluation involved experienced offenders at risk of jail or prison. Probationers assigned to HOPE were significantly less likely to produce positive drug tests or to be arrested over a twelve- month study period. These offenders spent about one- half as many days in prison on revocations or new convictions. (See table., reproduced from Hawken and Kleiman 00.) If HOPE were implemented on a wide scale, it might cut prison time substantially.. Incarceration and Drug Courts This section presents a new analysis of data on the incarcerated population, including both state prisons and local jails. First, we show that offenders with drug use problems continue to be a large share of those in jail and prison, and that recent entering cohorts of drug- using inmates are considerably older on average than late- s cohorts. Second, we assess whether under the usual eligibility rules, an expansion of drug courts could substantially reduce the numbers of drug users locked up... Data and Analytic Framework We analyzed two waves of data each from the Survey of Inmates in State Correctional Facilities (SISCF) and the Survey of Inmates in Local Jails (SILJ), comparing changes in the standing local jail and state prison populations from the latter part of the s to the early part of the 000s. Specifically, the prison data are drawn from the and 00 waves of the SISCF, and the jail data are drawn from the and 00 waves of the SILJ. For the prison population, our analyses focus only on state inmates since the federal inmate survey was only added in. Moreover, federal prisons account for fewer than percent of all those incarcerated on any given day. In both surveys, all data, including key indicators on prior offenses and on substance use, are based on inmate self- report. There is a substantial literature on such self- reporting in correctional settings both for criminal